Research Article
Performance of Post-Training Two-Bits Uniform and Layer-Wise Uniform Quantization for MNIST Dataset from the Perspective of Support Region Choice
Figure 7
Normalized histogram of weights (FP32) from (a) layer 1, (b) layer 2, and (c) layer 3; layer-adapted transfer characteristic of the symmetric two-bits UQ for Case 1 and (a) layer 1, (b) layer 2, and (c) layer 3; layer-wise FP32 and uniformly quantized weights from (a) layer 1, (b) layer 2, and (c) layer 3.
(a) |
(b) |
(c) |